AI

Mixup is a new, Mad Libs-style app for creating AI images from photos, text, and doodles

The ex-Googler team behind the 3D design app Rooms from Things, Inc. is out with his latest project: a playful AI photo editor called Mixture. The iOS-only app lets anyone create new AI-generated photos using “recipes,” which are Mad Lib-style fill-in prompts for your photos, texts, or sketches.

For example, you could ask Mixture to turn your scribbled sketch into a beautiful Renaissance painting, or to re-dress your pet in a fun Halloween costume. You can use a selfie to see what you would look like with a different hairstyle, or create something even crazier, like imagining a friend transformed into the form of Italian brain rot.

Image credits:Mixture

The app is built on top of Google’s Nano Banana, but the ‘recipe’ format opens up a new way to interact with the model: turning a generative tool into an online party game.

‘The thing that [Nano Banana] What no model did before was it could take your image and maintain it in a compelling way that wasn’t creepy,” said the founder and CEO of Things, Inc. Jason Toffwhose background includes working on experimental apps within Big Tech companies like Google and Meta, as well as time spent managing products at Twitter.

But what makes Mixup particularly fun to use is that the app’s “recipes” – the user-generated AI prompts – are shareable.

Image credits:Mixture

“Generative AI is so powerful, and yet most of the time you use these tools and think, here’s your text box: think of something creative. And what do you write?” said Toff, speaking about the shortcomings his team saw in the current AI landscape.

See also  Musk may still have a chance to thwart OpenAI's for-profit conversion

“And so, instead of having to be creative and think about what to create, you see something that worked, and you can just fill in the blanks,” he said.

Image credits:Mixture

Once users create a new prompt in Mixup, they can choose to publish it along with the resulting photo to a public feed or simply download it for personal use. From the feed, other users can view the photo and tap the button next to it for ‘Try Recipe’. This allows anyone in the app to reuse the recipe to generate an image with their own photo, text or doodles. (The latter can be done via a simple in-app drawing function.)

The team believes that seeing a photo next to the recipe used to create it could also help address the unpredictable nature of generative AI images.

“Another problem [with generative AI] is what we internally called a ‘slot machine’ problem, where you press the button, something comes out, you press it again, something else comes out, and you don’t feel like you have control over the output,” Toff explained.

Image credits:Mixture

However, in Mixup, users can see both the photo and the prompt used to create it in one place, giving them an idea of ​​what their output would look like. They can also toggle to see both the before and after images with a button, if the creator chooses to leave this setting enabled.

Additionally, similar to OpenAI’s AI video app Sora, users can upload their own photos to Mixup to use in their AI images. If you choose to do this, any person you follow in the app can also create AI images with your likeness – a feature it calls ‘mixables’.

See also  Photos of the Castaways that compete

The company hypothesizes that groups of friends would follow each other to take advantage of the feature, but a creator class could also potentially appear on the platform – if they don’t mind seeing themselves messed up in bizarre ways. (If you don’t want your image to appear there, don’t upload it or follow anyone.)

Image credits:Mixture

The app also uses OpenAI technology to address some of the expected moderation issues around AI images, but Toff admits that Mixup also relies heavily on Google’s built-in controls within its image model to limit things like sexual content or violence.

At launch, Mixup is optimized for iOS 26, but is supported on iOS 18 and later. If it is successful, a web version or Android app could be added later.

Free users receive 100 credits, which equates to $4. Meanwhile, images cost almost 4 cents to produce. When the credits run out, users can subscribe to different levels with 100, 250 or 500 credits per month.

The app will launch globally on the App Store at midnight on November 21, but will require an invite to enter. TechCrunch readers can use the code TCHCRH (until it runs out) to enter. Mixup can be pre-ordered before launch.

This post has been updated after publication to note that the app will launch on November 21st, not November 20th as previously stated.

Source link

Back to top button